Lugor

AI Overviews Has Arrived in France: Can Your Site Be Cited?

Illustration de l'article AI overview arrive en France

On June 29, 2026, Google sent a letter to French press publishers. Revealed by Ouest-France, it sets a timeline: AI Overviews and AI Mode will be rolled out in France before September 23, 2026. Both features arrive together, in French. France was the last major European market without them.

This is not an algorithm update. It is a change in what Google displays above your results.

What Will Appear Above Your Site

AI Overviews is an AI-generated summary placed at the top of the results page, answering the question before the links. AI Mode goes further: a conversational tab where search becomes a dialogue. Eight European Union countries have had them since March 2025, including Germany, Italy, Spain and Belgium. The United Kingdom has had them since summer 2024.

If France waited, it was not a technical choice. The 2019 neighboring rights law requires publishers to be paid when their content is reused, and the competition authority fined Google €250 million in March 2024 for training Gemini on press content without agreement. The June 29 letter therefore accompanies the rollout with three commitments: an opt-out right for publishers, separate performance indicators for classic search and AI search, and the extension of neighboring rights payments to content used in generated answers.

What the Markets Already Served Have Measured

Two years of data exist. They converge.

Ahrefs, across 300,000 keywords, measures the drop in clicks on the top organic result when an AI Overview appears: -34.5% in April 2025, -58% in December 2025. The Pew Research Center, using real browsing sessions, finds a click rate of 8% with an AI summary versus 15% without. Google disputes the methodology, but the order of magnitude matches the other studies. Seer Interactive, across 53 brands and 5.47 million queries tracked in 2025, puts the drop in organic CTR at -61% on the affected queries. Similarweb observes that the share of searches with no click at all rose from 56% to 69% in one year.

Two caveats apply to these figures, and they matter.

The first: the impact is asymmetric. Amsive, across 700,000 keywords, measures -19.98% on non-brand queries, but +18.68% on brand queries. AI Overviews trigger heavily on informational queries and very little on transactional ones. It is generic answer content that loses, not established brands, and not the pages people search for by name.

The second: being cited changes everything. Again according to Seer Interactive, a site cited in the AI Overview gets an organic CTR 35% higher than a site ranked in the same position but not cited. And the summaries do cite: three or more sources in 88% of cases according to Pew. Visibility does not disappear. It shifts toward the sites the machine chooses as sources.

What a Site Needs to Be Chosen as a Source

Google wrote it in its 2026 guide to AI search: there is no separate discipline for being visible to AI. It is still SEO, but SEO done seriously, not a plugin installed on redesign day.

In practice, four layers make the difference. None is a setting. On an existing site, each one is a project.

Complete structured data:

Google confirms it uses structured data for its generative AI features. A site where every page exposes its JSON-LD (Organization, Article, FAQPage, Service, Product) gives the machine an unambiguous reading of who is speaking and about what.

Controlled crawlability:

Generative engines are not limited to Googlebot. OpenAI, Anthropic, Perplexity and Mistral send their own bots. Look at what Google has just conceded to press publishers: an opt-out right, bot by bot. A professional site should have that control by default: explicitly deciding which AI bot accesses what, and tracking those that ignore the rules. On most sites, nobody even knows which AI bots read the pages.

Structured authority content:

AI summaries cite pages that answer precisely, with dated facts and a readable structure. A dense, well-organized page will be cited. Ten thin pages will be paraphrased without a link.

A baseline measured now:

Google promises separate indicators for classic search and AI search. They will only be useful if you know your numbers from before. Record your CTR by query group in Search Console this summer, because it is the only point of comparison you will have. And do not rely only on Google's measurement: your own site can tell an AI bot that is crawling from one that is fetching an answer for a user. That distinction is the difference between "AI reads me" and "AI brings me prospects."

That leaves emerging standards such as llms.txt, which describes a site to language models. Google has said it does not use it. But Google is no longer the only engine that reads the web: the standard also targets ChatGPT, Perplexity and Claude. Publishing it costs nothing for a well-built site; betting on it alone makes no sense.

Three Months: Catch Up on Your Site, or Start Over on Sound Foundations

The deadline is known: before September 23. There are two ways to get there.

  1. The first: upgrade what you have. Audit the structured data page by page, rewrite the robots.txt for AI bots, publish an llms.txt, rework the authority pages, retest performance. On a WordPress or classic agency site, each layer is a technical work package that means weeks of development, on a stack that was not designed for it, and to be redone with every change.
  2. The second: start from a system where all of this exists by design. Every Lugor site ships with structured data generated on each page: more than 20 typed schemas, enriched with the fields generative engines use, including summary and citations.
    A robots.txt that names and manages 21 AI bots individually (those of OpenAI, Anthropic, Perplexity, Google, Mistral and others), with the same power to accept or refuse that Google has just granted to press publishers.
    An llms.txt and its extended version, compliant with the standard, regenerated at every publication.
    Tracking of AI bot visits that distinguishes indexing crawls from requests triggered by a user: your own AI search indicators, without waiting for Google's.
    A multilingual sitemap in 5 languages, a perfect 100 in all four Lighthouse audits, and more than 400 criteria audited before delivery.

None of this was added for the occasion. It has been Lugor's architecture from day one: a site designed to be read, understood and cited by machines as much as by visitors. The June 29 announcement requires no upgrade from us because it will increase the visibility of what was already there.

Installation in under 48 hours. And the model follows the same logic as the promise: a license paid once, valid for life for your domain, with no subscription. You are not buying a compliance service to renew; you are buying an asset that is compliant.

The sites that get cited in September are the ones whose foundations are laid in July.

Key takeaways

  • AI Overviews and AI Mode will be rolled out in France before September 23, 2026, according to Google's June 29 letter to press publishers.
  • In the markets already served, organic clicks fall by 34 to 61% on queries with an AI summary, according to studies by Ahrefs, Pew and Seer Interactive.
  • The impact is asymmetric: −19.98% on non-brand queries, +18.68% on brand queries (Amsive).
  • A site cited in the AI Overview gets an organic CTR 35% higher at the same position: visibility shifts toward the cited sources.
  • Four layers determine citation: structured data, control of AI bots, authority content, measurement. They are built into every Lugor site.

Frequently asked questions

Glossary

AI Overviews
An AI-generated summary that Google displays at the top of the results page, before the organic links, citing the sources used.
AI Mode
A conversational tab in Google Search where search takes the form of a dialogue with AI, with no classic results page.
GEO (Generative Engine Optimization)
The set of practices that make a site readable, understandable and citable by generative search engines (Google AI, ChatGPT, Perplexity, Claude).
llms.txt
A standard file placed at the root of a site to describe its content to language models, complementing robots.txt.
Neighboring rights
French law (2019) requiring press publishers to be paid when their content is reused, notably by platforms and AI.

Boost your visibility

Be visible and cited by AI
with Lugor.